Summary: | 碩士 === 國立成功大學 === 統計學系 === 89 === Process capability indices (PCIs) has been widely used in industry today. The potential performance of a manufacturing process can be measured and compared by PCIs. PCIs can provide direction for quality improvement. In most cases, the measure-ments are approximately normally distribution with symmetric tolerances. In some cases, process may have asymmetric tolerances. Boyles (1994) pointed that the classical PCIs
(Cp, Cpk, Cpm and Cpmk) can not reflect the actual process capability very well for asymmetric tolerances. This paper tries to provide a better PCI for asymmetric tolerances.
Under normality and using the concept of desirability function of response surface methodology (RSM), we propose
a new class of process capability indices Cpm(s,t) which is modified from Cpm to deal with the problems when tolerances
are asymmetric. By comparing several statistical properties (such as the power, mean square error) of the natural estimator , we recommend combination of for asymmetric and symmetric tolerances.
In chapter 4 and 5, we compare the recommended Cpm(s, t) with several indices which is proposed for asymmetric toler-ances in recent years. The results show that the performance
of the recommended Cpm(s, t) is very well. Finally, an example of real data analysis is provided to illustrate how to use the recommended Cpm(s, t) to evaluate manufacturing.
|